Here are some of my notes in relation to assessing a site when it comes to VMWare and visualization in general.


The following items need to be taken into consideration in relation to CPU when it comes to visualization.

  • CPU Speed (in MHz)
  • Usage (Processor(_Total\% Processor Time) – The % of processor being used.
  • Queue (System\Processor Queue Length) – The number of threads that are ready for execution but are being held in a queue because the processor is too busy (More then 2 generally indicates CPU congestion)

The CPU needs to be converted into MHz to be useful. For example a Server with 1 x 3GHz CPU running at 10% would be 300MHz of CPU usage. Another example. A server with 4 x 2.7GHz CPU @ 16% would be using 1728MHz

VMWare Limit – 32 CPUs

CPU Considerations

  • Application processing is the least CPU overhead on the host, Disk and network increase CPU overhead for the host.
  • The higher CPU, the higher the overhead.
  • It is also dependant upon the Guest OS.

When looking at existing servers take into account

  • Sockets
  • Cores
  • Hyper Threading
  • Speed


The following needs to be considered when looking at memory.

  • Available Memory (Memory\Available Bytes) – Percentage of available Memory
  • Page file Usage (Paging File (_Total)\% Usage) – % of allocated page file being used
  • Paging (Memory\Pages/sec) – Number of paging operations per second (to the disk)
  • File Cache – Memory reserved for file system cache

VMWare Limit – 1TB

Memory Considerations

  • Same OS and applications = saved memory due to transparent page sharing (deduplication)
  • If all of the machines are similar OS and applications savings of 30% can be achieved.

When looking at existing servers take into account

  • Size of memory
  • Page File


The following needs to be considered when looking at disk.

  • Disk IOPS (Physical Disk(_Total)\Disk Bytes/sec) – Rate of read and write operations on disk. This much factor in the overhead for RAID operations
  • I/O Speed (Physical Disk (_Total)\Disk Transfers/sec) – Rate that bytes are transferred to/from the disk during write/read operations (AKA Transfer Rate i.e. 100MB/Sec)

When considering the required storage you need to take into account the swap file that is created on boot (Same size as memory, so 32GB for a serve with 32GB memory) and Snapshots.

VMWare Limit – 2TB per HDD

When looking at existing servers take into account

  • Current HDD size in MB
  • If the OS and Data on a machine is split across Drives
  • IOPS that are in use by the machine
  • Type of disk


The following needs to be considered when looking at network.

  • Speed (Server\Bytes Total/Sec) – number of bytes the server has sent to and from the network

VMWare Limit – 10 NICs

When looking at existing servers take into account

  • Number of NICs
  • Speed of NICs
  • Net performance data (i.e. how many packets sent and received)

Load Profiles.

Need to take into consideration the Peak Usage. Not just machine averages. If peak usage is not considered then during heavy loads the users will experience poor performance.

Working out Host Peak Capacity

When estimating capacity, you may only wish to load a server to 80% for example to accommodate for spikes.

Estimating Required Capacity Rule of Thumbs

  • Roughly 4 – 5 VMs per Core per Host assuming Sufficient Memory, network bandwidth and machines are single CPU machines

Costing / TCO / ROI

Direct server costs.

  • Hardware cost
  • Support Contracts
  • Third Party software and support
  • Networking costs – Additional Switches, cables etc
  • SAN costs – additional storage / FC Switches

Indirect costs

  • Data center cooling and power
  • Server administration costs
  • Server provisioning costs

ROI calculation
Return on Investment (ROI) = (Returned Value – Initial Value) / Initial Value